XPath

What it does

First, during the process of creating the roadmap, the Gemini API is used to generate an outline for the course based on the topics users search for. For each content item in the outline, we use the Google Search API to search for resources (text, video) combined with certain ranking algorithms to suggest materials to users.

Additionally, the Gemini API plays an important role in automatically generating questions based on the lesson content. To address the hallucination issue of large language models (LLMs), we employ Retrieval-Augmented Generation (RAG) to retrieve information relevant to this lesson.

Finally, to create personalization, we store the test results for each lesson and use them as data for Gemini to analyze the user's learning process, including identifying areas where the user has not fully grasped the material and recommending review sessions.